|
PureBytes Links
Trading Reference Links
|
Yarroll asked:
> Please have a look at the formulas:
>
> Buy long: divergence(C,LinRegSlope(C,2),0.01)=-1
> Sell short: divergence(C,LinRegSlope(C,2),0.01)=1
>
> They are very profitable in tests, just too good to be true. That's
because
> it is... Tests results back-adjust themselves (etc.)
This reminds me of a problem I ran into a while back. Reasoning that a peak
or trough, once established, ought to remain a peak or trough, I decided to
see whether one could make a profit by taking action after the change of
direction had been recognized. So I wrote a system that said, in effect,
"Take a position after the price has moved X points down from the peak
or up from the trough," where X was the size of the move required to
establish
that a peak or trough had been made. Worked like a charm! Except, of
course, that when I looked at it on short intraday bars, about once a week
it
would decide that a bunch of peaks and troughs had never happened, and
the up or down trend had been continuously in effect since almost forever.
If I'd actually been trading, significant money would have gone the wrong
direction.
So, fine. Now I accept that there is no way around the peak/trough problem.
Yet I can't help thinking that a peak (or trough) established by prices X
percent
lower (or higher) on either side ought to remain a peak or trough, no matter
what else happens later. Can anyone explain what it is about the way the
relevant
functions are written that invalidates this reasoning? Or direct me to a
resource
that will explain it to me, in very small words?
Many thanks.
Owen Davies
|